In this study, we investigated the role of TGF-β in a well-defined cohort of early-stage, node-negative invasive breast cancer patients who were part of the SweBCG91RT randomized radiotherapy trial. We used phosphorylated SMAD2 as a surrogate marker to assess active TGF-β signaling. TGF-β is known to have both tumor-promoting as well as restraining functions, depending on tumor type, tumor characteristics as well as the status of the tumor microenvironment [10,11,12,13]. The role of TGF-β in breast cancer is therefore still not fully understood, posing challenges to the implementation of TGF-β targeting therapies, despite available and specific inhibitors such as galunisertib or fresolimumab [22].

We found that 45% of early-stage breast cancer patients exhibited high pSMAD2 levels, defined as having ≥ 80% of positive tumor cells. Interestingly, patients with medium pSMAD2 had a significantly increased IBTR as well as any recurrence risk compared to the pSMAD2 high group.

These results may suggest that, in early-stage breast cancers, TGF-β signaling still exhibits its tumor-suppressive, anti-mitogenic function, thereby limiting progression and lowering recurrence risk. Preclinical studies suggest, that during tumor progression, key components of the canonical TGF-β pathway (e.g., SMAD2/3, SMAD4, TGFBR1/2) are often epigenetically downregulated via histone modifications, potentially leading to non-canonical pathway activation with pro-tumorigenic signaling routes [46, 47]. Consistent with this, we found that tumors with medium or low nuclear pSMAD2 were larger and of higher grade than those with high pSMAD2. Importantly, even after adjusting for standard clinicopathological factors, medium pSMAD2 remained an independent predictor of increased recurrence risk, indicating that attenuation of TGF-β signaling may serve as a prognostic marker at early breast cancer stages. Of note though, no significant differences were noted between the pSMAD2 low group and the medium or high groups with regard to IBTR or overall recurrence risk. This patient group may represent a distinct subset with unique downstream signaling. However, since pSMAD2 low patients comprise only 17% of our cohort, this finding should be interpreted cautiously.

Future studies incorporating mutational and expression data on TGF-β pathway components are needed to improve patient stratification and therapeutic targeting. Unfortunately, in our early-stage breast cancer cohort, no information was available regarding the mutational status or gene expression level of TGF-β signaling components. The value of integrating TGF-β downstream signaling elements in breast cancer biomarker studies was highlighted in a study on invasive breast cancer, which included a substantial number of node-positive patients [38]. This study demonstrated that protein levels of the tumor suppressor SMAD4 could further stratify the pSMAD2 high patient group, revealing significantly reduced relapse-free survival rates for pSMAD2 high/SMAD4 low patients compared to pSMAD2 high/SMAD4 high patients. This effect may be attributed to the onset of non-canonical TGF-β signaling in the absence of SMAD4, impairing the anti-proliferative action of TGF-β [48]. Our future efforts should consider to integrate SMAD4 into our pSMAD2 analyses, ideally, by investigating their direct interaction.

Based on our findings, targeting TGF-β in early-stage breast cancer patients without appropriate biomarker guidance may be unfavorable. While our data suggests pSMAD2 as a potential biomarker, we acknowledge the technical challenges and limitations associated with our study. Tumoral nuclear pSMAD2 positivity was scored by a board-certified pathologist specializing in breast cancer pathology using a predefined staining cut-off designed for clinical relevance. Future efforts should focus on developing a standardized scoring system for pSMAD2 IHC. pSMAD2 may transiently localize in the cytoplasm, and consequently, in numerous cases a significant background cytoplasmic staining was observed, complicating the interpretation of nuclear staining. Digital approaches are essential to standardize and ensure the transferability of pSMAD2 assessment, but despite significant advancements in AI-based tissue annotation tools, the accurate digital discrimination between tumor and stroma cells based solely on hematoxylin staining remains problematic. Multiplex fluorescent approaches, incorporating specific markers for tumor and stroma cells, could enhance the digital assessment of tumoral pSMAD2 and should be considered; yet, as for now, multiplex approaches are difficult to integrate into routine clinical diagnostics. Another important limitation of our study is the low number of HER2+ and triple-negative cases, which prevents meaningful subtype-specific analyses. That said, the subtype distribution of our randomized trial cohort reflects reported distribution of early-stage breast cancer [49, 50], and we observed no statistically significant difference in pSMAD2 levels between subtypes (Table 1). Given the generally favorable prognosis and low recurrence rates in early-stage breast cancers, we decided not to split the cohort into the individual subtypes for recurrence risk analyses to preserve statistical power. Consequently, our findings are driven predominantly by Luminal tumors (84.8%) and offer very limited insight into early-stage HER2+ and triple-negative breast cancers. This gap warrants dedicated studies in these subgroups, especially since analyses of later-stage disease have revealed subtype-specific effects, such as HER2-driven SMAD3 phosphorylation, and our group and others have demonstrated that the tumor microenvironment carries distinct molecular characteristics and clinical value in the different subtypes [42, 51, 52].

Indeed, our analysis showed that the inverse relationship between tumoral pSMAD2 levels and overall TILs density was confined to the Luminal subtypes and was not present in the pooled HER2+/triple-negative patients. Furthermore, only Luminal B tumors displayed that higher pSMAD2 correlate with reduced FOXP3 and PD-L1 expression. Our earlier work on the same cohort demonstrated that total TILs numbers, and thereby also specific subsets including CD8⁺ and FOXP3⁺ cells, rise progressively from Luminal A through Luminal B to HER2⁺ and Triple-negative subtypes [42, 43]. Such subtype-specific immune landscapes imply that TGF-β signaling also exerts its immunomodulatory effects in concert with tumor intrinsic pathways, leading to differential immune activation or suppression across subtypes. This again highlights the context-specific nature of TGF-β signaling and underscores the importance of well-defined patient groups for determining the optimal therapeutical setting.

The immunomodulatory effects of TGF-β not only indirectly affect tumor progression but also treatment responses. RT generates “danger” signals in tissues that, under certain conditions, enhance tumor antigen presentation and thus anti-cancer immunity [14, 53]. A parallel inhibition of TGF-β signaling during RT is therefore expected to enhance RT efficacy by promoting immune cell activation [54]. Indeed, findings from a randomized trial in metastatic breast cancer, where patients receiving a high dose of the TGF-β1,2,3 neutralizing antibody fresolimumab experienced prolonged overall survival compared to the lower dose group, support a role of TGF-β signaling in RT response [55]. Our retrospective analysis focused on lymph node-negative, early-stage breast cancer patients does not indicate that the increased IBTR or any recurrence risk observed in the pSMAD2 medium patient group is due to differential effects of adjuvant RT based on TGF-β signaling in the primary tumor, as proven by the consistent and significant efficacy of RT across all pSMAD2 patient groups, with a non-significant formal interaction test between pSMAD2 group and RT.

In conclusion, using nuclear pSMAD2 as a marker for active TGF-β signaling in a large cohort of early-stage breast cancer patients, our findings suggest that TGF-β signaling may play an anti-tumoral role by reducing recurrence risk, challenging its therapeutic targeting in this patient group. We observed no interaction between TGF-β status and benefit from adjuvant radiotherapy. In Luminal subtypes, higher pSMAD2 correlated inversely with TIL density, highlighting subtype-specific immunomodulatory effects. Moving forward, it is crucial to standardize pSMAD2 assessment methods and to expand analyses to larger HER2+ and Triple-negative cohorts of early-stage breast cancer to fully decipher subtype-specific roles of TGF-β signaling.